I read the draft once again and I'm a bit more confused than before...
Here are a few comments:
• The document should clarify what "ERX" means and how ERX relates to ERP.
• The first message of the protocol sequence in Sec. 3 ends with a
comma. Is it missing a notification?
• "EAP_Initiate/Re-auth message replaces the IDi payload": Omitting IDi
may or may not be the right decision (it modifies RFC 5996!), but in any
case a rationale is needed. In order to not break IKEv2, the protocol
could mandate that the same ID should be used at both levels, i.e. that
IDi should be identical to the EAP-level identity. I realize that this
is not ideal.
• The first paragraph of 3.2 is confusing: most deployments would
probably use a single EAP server *even if* they use multiple IKE/IPsec
gateways. Hence the last sentence of the paragraph should be dropped.
• More importantly: the issue of identifying the client (e.g. for the
gateway to generate useful audit records) is important even for the
"pure enterprise" use case of roaming from 802.11 to IPsec.
Unfortunately the protocol is incomplete until/unless any of the methods
defined in this section (e.g. transmitting the client's authorization
group) is standardized. This is probably just a question of picking the
right RADIUS TLV. Maybe just need to say that the Class attribute MUST
be used. In other words, in the absence of IDi the gateway cannot make
policy decisions even if only a single gateway is used!
• Sec. 4: how is the realm encoded? I think the correct answer is
"ASCII, with no null termination".
Thanks,
Yaron
On 11/20/2011 10:01 AM, Yoav Nir wrote:
Hi Yaron
Actually the motivation in my case is a smooth transition from a
802.1x local network, to remote access VPN on a 3GPP/WiMax public
network and back, and this is a very enterprise network sort of thing.
At the HOKEY meeting in QC there were some Telco people, and they
didn't seem to think there was another use case.
I do remember the use case of doing IKE with EAP-SIM or EAP-AKA, but
IIRC that was also the phone connecting to its home network over the
Internet.
Qin: are you aware of cases where IKE is used with anything other than
the home network?
Yoav
On Nov 20, 2011, at 9:42 AM, Yaron Sheffer wrote:
Hi Yoav,
motivation for this work seems to have come from 3GPP/3GPP2/WiMAX,
and I strongly suggest that you or your coauthor go back to the
originating organization to validate your use case(s).
I find the new paragraph (top of Sec. 3.2) confusing: I would expect
the IKE negotiation to go to a local network (in the "visited
network") with this gateway being supported by a "home" EAP server.
EAP requests are commonly routed back into the home network. In a
telco network, this backend EAP connectivity most likely would *not*
be over the open Internet.
Lastly, judging by the level of interest so far, I do not see this
draft becoming an ipsecme WG charter item. I do not have any problem
with its being published elsewhere.
Thanks,
Yaron
On 11/19/2011 02:07 PM, Yoav Nir wrote:
On Aug 6, 2011, at 10:37 PM, Yoav Nir wrote:
Hi
At the meeting in Quebec, I gave a presentation at the hokey meeting
abouthttp://tools.ietf.org/html/draft-nir-ipsecme-erx .
The draft covers using the EAP extensions for re-authentication in IKEv2. The
obvious (to me) use-case is a phone connected to a 802.1x network. As you leave
the building, the same phone automatically using IKEv2 over a 3G network
without the user authenticating, by using the handed-over keys from 802.1x.
ERP (RFC 5296) works in two cases:
1. when the new AAA backend and the old AAA backend are the same, and
2. when they are different - you connect to a local EAP server
There is an open question here. Obviously, when you use EAP for 802.1x or PPP or some
other network access, you often connect to a local Authenticator that is not the same as
your "home network". But is this relevant in IKEv2? IKEv2 is used over the
Internet. Why would you ever want to connect to a server other than your home (or a
server that relies on the same AAA backend)
In other words: is there a use-case for connecting to a local rather than a
home server in IKE, a use-case that uses EAP.
My feeling is that the answer is no, and there were some phone operators in the
room who agreed with me. Someone did bring up the case of host-to-host IPsec,
but I don't think that ever uses EAP.
Does anybody have different thoughts about this?
(crickets)
As there were no replies to this email, and as there was pretty much an uncalled
consensus at the HOKEY meeting, I have submitted version -02 of the draft with an extra
paragraph in section 3.2 to explain that "roaming to a different EAP server"
scenario is probably not relevant.
http://www.ietf.org/internet-drafts/draft-nir-ipsecme-erx-02
I would be happy for this to become a working group item, but if not, I would
like to take it to our ADs (not sure which one, as this involves both IPsecME
and HOKEY). I would also appreciate any suggestions for the Security
Considerations section, other than just moving the rest of section 3.2 into it.
Yoav
_______________________________________________
IPsec mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/ipsec
_______________________________________________
IPsec mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/ipsec
